Algorithms Research And Realization For Components Optical Detection Systems

With the rapid development of electronic information industry, the demand forelectronic products is growing, which greatly promotes the development ofmicro-electronic components of the technology and market demand, and also makes thedemand for manufacturing equipment of micro-electronic components increasing. Thechip components accountes for most of micro-electronic components production. Theproduction of chip components is a very complex process, which is a high degree ofautomation and needs high dependence on the production equipments. High speedtaping machine is one of the most important equipments in chip components production.It's main function is to classify and transfer the small piece of chip components to aplate automatically rapidly and orderly in accordance with the requirements, andremoves the electronic components with bad performance parameter and appearance.The task of removing the bad components is just accomplished by automated opticalinspection (AOI) system of the taping machine. However, so far there is a great gap inhigh speed taping machine area between our country and developed countries.In view of this, the importance of high speed taping machine manufacture isobvious. The main work of this thesis which is against the background of project "HighSpeed Taping Machine Development and Industrialization" focuses on the algorithms'study and realization. The main work in this thesis can be summarized as following:1. Study of the image acquisition problem for the AOI hardware system. The studyof the image acquisition system makes the development of AOI software not only stopat the theoretical level, but also benefit the algorithms tests in the sequel. In the desighof image acquisition system, a creative structure of the turntable device is proposed,which makes the continuous image acquisition possible. Turntable device can acquisit alarge number of image data in a short time, which helps greatly for the stability,robustness and universality test of the edge detection algorithms, and classification ofdefective component and mode establishment in defect detection algorithms tests.2. Realization of the image straight algorithm. The projection method is applied tofind the geometric axis of the image, then this thesis uses rotation algorithm to completethe image straight. Finally the selection of the step angle which impacts on the imagestraight accuracy is analyzed, and it is shown that the image straight accuracy changeswith the selection of step angular size, specifically, the smaller the step angle is, thehigher the image straight accuracy is. But, the processing time will increase with thedecrease of the step angular.3. Realization of the edge detection algorithm and improvement of some relatedalgorithms. Firstly, several algorithms are chosen and implemented according to projectrequirements, and then some appropriate algorithms are improved, thus a desired edge detection result is obtained. Comparing three edge detection algorithms, it is shown thatSobel algorithm can get better image edge, but is not good at noise attenuation. Thus, animprovement is made to the Sobel algorithm, and the noise attenuation ability can begotten a good promotion. Canny algorithm can obtain more sharp edges, but the noiseattenuation ability is shown to be poor. In addition, Canny algorithm is complicated thusits realtime property is poor. In dealing with the edge detection, the wavelet transformalgorithm is basically equivalent to the Canny edge detection algorithm. Moreover, inthe noise attenuation, the wavelet transform algorithm is as good as improved Sobelalgorithm. Therefore, the wavelet transform algorithm successfully bridges the gapbetween the edge detection accuracy and noise attenuation. However, the wavelettransform algorithm is the most complicated in the above-mentioned three algorithms,thus the realtime property is worst.4. Study of the corner detection and the defect detection algorithms, theirrealization with Hexsight. Considering the realtime property of the algorithm, the cornerof the matching algorithm is chosen to detect defects in chip components, and Hexsightsoftware is used to realize the defect detection, and a perfect test result is obtained. Forthe corner detection, the traditional Harris corner detection algorithm is applied andrealized by Hexsight software too, a good test result is also obtained.